Do You Guys Want to Dance: Zero-Shot Compositional Human Dance Generation with Multiple Persons

TL;DR

This paper introduces a new task, dataset, and evaluation protocol for compositional human dance generation involving multiple persons and backgrounds, revealing current models' limitations and proposing a zero-shot framework to improve realism and consistency.

Contribution

The paper presents MultiDance-Zero, a novel zero-shot framework for multi-person dance video synthesis, including pose-aware inversion, compositional augmentation, and consistency-guided sampling.

Findings

Existing methods fail to generalize to real-world multi-person scenarios.

MultiDance-Zero significantly improves realism and temporal consistency.

The approach outperforms state-of-the-art methods on the new cHDG benchmark.

Abstract

Human dance generation (HDG) aims to synthesize realistic videos from images and sequences of driving poses. Despite great success, existing methods are limited to generating videos of a single person with specific backgrounds, while the generalizability for real-world scenarios with multiple persons and complex backgrounds remains unclear. To systematically measure the generalizability of HDG models, we introduce a new task, dataset, and evaluation protocol of compositional human dance generation (cHDG). Evaluating the state-of-the-art methods on cHDG, we empirically find that they fail to generalize to real-world scenarios. To tackle the issue, we propose a novel zero-shot framework, dubbed MultiDance-Zero, that can synthesize videos consistent with arbitrary multiple persons and background while precisely following the driving poses. Specifically, in contrast to straightforward DDIM or null-text inversion, we first present a pose-aware inversion method to obtain the noisy latent code and initialization text embeddings, which can accurately reconstruct the composed reference image. Since directly generating videos from them will lead to severe appearance inconsistency, we propose a compositional augmentation strategy to generate augmented images and utilize them to optimize a set of generalizable text embeddings. In addition, consistency-guided sampling is elaborated to encourage the background and keypoints of the estimated clean image at each reverse step to be close to those of the reference image, further improving the temporal consistency of generated videos. Extensive qualitative and quantitative results demonstrate the effectiveness and superiority of our approach.